VITALIS: A Patient-Owned Agentic Framework for Emergency Detection and Governed Clinical Disclosure

Anjali Sharma-Tiwari (1), Priyank Sharma-Tiwari (1), Raeyaan Valleru Muppaneni (2), Rishon Valleru Muppaneni (2), Rohan Kalavar (3), Nirvik Kasula (4), Joshua Amaladass (5), Alex Colin (5)

(1) Basis Independent Silicon Valley, (2) Irvington High School, (3) California High School, (4) Dougherty Valley High School, (5) Robolabs

Abstract— Emergency response for a person living alone with chronic illness rarely fails at detection. Consumer wearables already detect falls and irregular rhythms, place emergency calls without user input, and, where users opt in, deliver a static medical profile to dispatchers. What they do not establish is why an alert fired. Detection operates as a binary trigger evaluated against population-calibrated thresholds, with a cancellation window as its only correction, and the resulting false-alarm burden is measurable in dispatch centers. Personal baselines narrow the problem without dissolving it, since a baseline-corrected detector can establish that a signal moved but not what moved it. Meanwhile prehospital clinicians report persistent difficulty accessing patient health information, and community responders to automated alerts name missing information about the cause of the alert among their principal concerns. Among the deployed systems and literature reviewed for this proposal, none interrogates an ambiguous signal before acting on it, and none delivers a governed clinical account to whoever responds.

This paper presents VITALIS (Vitals Interpretation, Triage, Agentic Local Intelligence System), a framework proposed for that gap. Instruments the person already owns report to an agent hosted on patient-owned hardware, a home hub with a smartphone as its mobile extension, which maintains a contextually conditioned characterization of that individual's ordinary physiology. Deviations are banded by magnitude and source confidence: expected variation is ignored, unambiguous time-critical signals escalate at once, and the ambiguous middle enters a resolution stage before any response is selected. That stage is the distinguishing contribution. It queries the wearer, requests a confirmatory reading from a higher-confidence instrument, and extends the observation window, bounded by a time budget scaled inversely to assessed severity, with unresolved states escalating rather than resting. Response then proceeds through a graded ladder in which each disclosure is scoped by the recipient's role, limited in time, and recorded, extending established break-the-glass discipline past the institutional boundary at which it currently stops. The agent reports observations and their provenance rather than clinical conclusions, leaving diagnosis with the clinician. VITALIS is presented as an architectural proposal, developed to the point where its architecture, operating logic, and limits can be examined and built upon. If developed as intended, it could give a responder arriving at a residence the account a conscious patient would have supplied.

Full Article: Sharma-Tiwari, A., et al.

Keywords: Chronic Disease Management, Wearable Sensors, Patient Monitoring, Emergency Medical Services, Intelligent Agents, Edge Computing, Personal Health Records, Health Information Privacy